Here's a quote from Michael's 1DS2 review:

"When the 11 Megapixel 1Ds appeared I expressed concern that it could well be challenging Canon's best lenses in terms of resolving power. I could clearly see differences when using some of my prime lenses, vs. several of the more popular zooms that we all use (including some L glass). Now with the 16 Megapixel 1Ds Mark II I believe that this is even more the case. "

I see this sentiment echoed more and more now that folks are looking to the 1DS3 expected at Photokina - presumably with more (22?) resolution than the 1DS2.

Here's what I think so far - but mostly questions - and I'd appreciate some comments to either elaborate or steer me in the right direction:

The problem seems most apparent with the Canon's wide angles - presumably the 16-35 or 17-40, (and wider primes as well?) but there also seems to be some concern with the 24-105 in a full frame camera (not surprising given the MTF chart). I haven't seen complaints regarding the other "work-horse" zooms - the 24-70 or 70-200. And these problems seem limited to the edges - not many complaints from the 1DMK2 users? (like me). As an aside, where would you draw the line on an MTF chart to reflect a 1.3 sensor? I recall that a sigma 14mm was actually not too bad on my d30 (back in the olden days).

Is diffraction expected to be an exacerbating issue? At ff 22mpx the pitch is the same as the existing 30d and I don't recollect seeing a lot of complaints re lens issues on that camera (but I don't follow that model very closely). Sure the photo sites are a bit smaller than the 1ds2 (.064 vs .072) but the diffraction impact is what it is based on the aperture vis a vis the pixel size - so diffraction shouldn't really be characterized as a lens challenge kind of problem should it? But it will certainly affect overall system resolution.

Taking "resolving power" as the interaction of all parts of the system - what needs to be considered in determining the degree of "challenge" to the image captured (ignoring capture technique - ie tripod, mlu, delay etc)?

How do you integrate MTF performance with sensor resolution (assuming ff)? Ie how do the line pair measurements at 10 and 30 lp/mm relate to pixel size? Or are the pixels at .064 small enough that this isn't a relevant consideration? Bottom line question is - are there any particular considerations you need to keep in mind when looking at an MTF chart in the context of the hypothetical 1DS3 - or is it relevant to the lens only? (Here I apologize since I vaguely recollect this haveing been discussed previously).

To what extent would a 1DS3 actually perform more like a lower resolution sensor in terms of overall resolving power given the more problematic lenses? Or even lower yet at higher apertures where diffraction starts to be a factor?

And then what happens when we take into consideration capture technique - eg: what happens to the 1/fl rule for handholding (ignoring IS)? Does it need to be 1/2xfl or 1/3xfl to make sure that ultimate resolution isn't limited by shooting technique? Presumably this issue isn't limited to ff, but any hi resolution camera. If that assumption is right then what's the implication for hi res, small sensor P&S's eg: Canon 8 mpx Powershot S80? In a similar vein, at what shutter speeds could you effectively hand-hold a camera with a P45 back for any given fl?

While there are other reasons to buy a "1" body, I wonder if the full value of 22mpx is going to be achievable by anything other than a relatively small handful of lenses (maybe third party), using a narrow range of apertures, tripod, mlu and delay.

Quite a mouthful of questions and queries... Let me add my own..

As a general trend Camera makers this year seem to have stepped off the pedal on the megapixel wars, rather adding more realworld features like IS/OS/VR to lenses, digicams and the like, rather than packing more pixels into the same finite space.. Unless there is some drastic revolutiion in sensors it's safe to say that we won't push the mp envelope for a while.

As for canon and it's widely expected 22 FF camera at photokina... this can only exacebate the issue of lens resolving power. As the 1 Series grows in pixels it's usefullness will actually decline, making it a specialist product, as well as the premier camera to lust after for us mere mortals that can barely afford the (very wonderful) 5D. Adding resolution will mean this camera will become a medium format tool.. and will not be as useful as the 35mm format always has been with the qualities of being small, light, with a full range of lenses for all conceivable situations.

I expect that Canon will find it will have to concentrate on newer lens designs very soon.. There is a vast sellection of substandard L glass out there that is in serious need of an upgrade. Some designs are hand-me downs from the breechlock mount days, with the register simply lenghtened for the longer EF mount as well as addition of the USM motor. The "new" 85mm f1.2 is essentially the same lens as in 1980 odd when it came out. Not a tremendous amount of R&D for this version "II", which didn't even get a rubber lens mount or a standard 77mm lens thread!
As we climb into medium and large format image quality these (substandard, not the 85mm lens I just mentioned which is excellent) lenses will surely show up their faults, and Canon will, I hope, respond with something new.

Then again they are an optical company so they are all for making new optical products to offer customers. This new 1Ds3 22mp camera will oblige users to "upgrade" to these newer lens designs with higher resolving power. Some lenses will always suffer, such as the ultrawide zoom family as well as the DO optics. Will this mean a resurgance of primes (and expensive ones at that!).

Probably not, as most consumers and pros are interested in convenience. Lugging around a few primes when one zoom will do makes sense from almost all perspectives (except image quality), as weight and bulk are an issue!

So Canon will bring out their megapixel wonder, but will also I belive cater to the mere mortals out there by bringing out other FF sensors without the megapixel punch. Effectively 22mp should be the limit for 35mm digital.

As we're talking of FF digital sensors I'd like to renew my hope that a Foveon-like solution is in the works, so that we can move away from the Bayer pattern sensors with all their associated problems of filters which effectively defocus the image in order to get away from moiré. If these problems could be "licked" even in the bayer pattern sensors then you would truly see a jump in the true resolving power of the sensor. This will mean an even bigger divide for the present lenses. You certainly wouldn't need as many pixel sites to capture the same detail. Patent issues may prove a bit of a problem for someone like Canon though. Nikon, seeing as they're shopping around for sensors might be the first to implement this type of solution for a full frame camera.
Fundamentally we are asking a 35mm camera to do a Medium format and Large Format job. I think it's fair to say that somewhere along the line compromises are going to be made, be it lens resolving power or the usability of these newer cameras.

So, sorry if I've only added more questions to yours.. all this can be answered once the Photokina curtain comes down (or by a few "leaks" a few weeks before!).

Till then, happy shooting...

Even with a 5D lenses are a huge issue. Only the best primes at optimum appertures get the best out of the sensor.

My hit on it is that folks are happy with the resolution of the of the 20D/30D, and for that matter Nikon D2X owners love the enhanced resolution too. So at least on center of EF lenses or on some EFs lenses the resolution is good, and diffraction limitations of the existing 30D sensor should apply. I haven't seen folks grousing about diffraction limitations with these cameras, they learn to live within these constraints. On a 22+ mp 1DSIII the challenge of the lens will be to maintain that level of performance the edge of the frame. Of course there is some performance fall off, especially at wide angle. Canon owners have been complaining about SOTA WA designs for quite some time, and I have read that Canon does intend on spending more effort in supplying this market. The ebay prices of approx $3500 for Zeiss Distagon 21 has not escaped Zeiss nor Canon!
Mike K

22MP would be helpful to Michael and other landscape photographers who can use low ISO and shoot with a tripod, the best lenses and mirror lockup, but it is most likely overkill for most hand held 35 mm type work. The above measures would be necessary to really make use of the increased resolution. The extra MP would at the expense of less dynamic range and more noise due to the smaller pixel size. Indeed, Michael and others on his recent African expedition noted that the 5D gave smoother renderings with higher ISOs.

The 22MP camera would be more of a specialty item. From what I understand, the EOS 1D Mark II outsells the 1Ds Mark II by a large margin. The former could be updated to full frame at 12MP with the same pixel size as before and this resolution would be sufficient for most needs.

I would'nt be surprised if canon came out with a new mount, just for the higher rez full frame cameras... probably put a chrome rim on it to!

witzke-studio.com

There seems to be no doubt that a 22mp full frame could be capable of providing marginally higher resolution in the central part of the image which is captured by the crop format cameras such as the 20D, 30D and Nikon's D2X. There have been so many comparisons on the internet which demonstrate the slightly higher resolving power of the 'cropped format' sensors with higher pixel density. The 1Ds2 has approx. the same pixel density as the D60 and 10D and any future 22mp successor to the 1Ds2 would have approx. the same pixel density as the current D30. The Nikon D2X with its 1.5 crop factor already has greater pixel density than a 22mp full frame.

We seem to have a situation where the D2X is considered to produce images at least as sharp as the old 1Ds and only a whisker behind the 1Ds2. That fact in itself would suggest that there are at least some lenses that will be up to the job of providing greater detail with a 22mp sensor. However, I would have to agree that for edge and corner resolution, 22mp would more often than not, be overkill.

But we should get this into perspective. Generally, edge and corner resolution do not carry the same weight as the central part of the image. Photodo ratings take this into consideration. They are 'weighted' ratings and a lens is not marked down as much for poor edge performance as it is for poor performance towards the centre. This policy of Photodo is presumably in line with common perception. Furthermore, I would suggest that photographers using shallow DoF in a creative way will often not be concerned at all about resolution fall-off in the corners.

Wide angle lenses are going to present a problem, but they already are a problem. When I got my 5D, I was a bit concerned that edge resolution at 15mm with my Sigma 15-30 would be a disappointment. Surprisingly, it wasn't. Nor is vignetting a problem at the small apertures I use for most landscape and architectural shots. What is much more noticeable than any resolution fall-off is the pronounced distortion towards the edges of the frame. This is a disaster if there's a person near the edge. Their body and/or head becomes very noticeably misshapen.

Bjanes has a point that greater pixel density will require the use of faster shutter speeds to make full use of those extra pixels. But this not as great a problem as one might think. A question that often cropped up in the past, on this forum, when people were struggling to get used to the implications of the 'cropped' format, was 'how is the 1/FL rule affected?' 'If I've been using 1/100th sec with a 100mm lens on my 35mm film camera (for a reasonably sharp hand-held image), should I use the same shutter speed with my 10D?' The answer was generally, no. The rule should be amended to 1/FL*1.6. That is, with a 100mm lens, one should use a shutter speed of 160th for a reasonably sharp hand-held shot, assuming we're talking about the same FoV and same print enlargement.

Of course, whether 1/FL is a good guide or not is another issue. Whatever your standards are for a sharp hand-held shot using a 1Ds2, then they would have to be increased by a relatively small margin. Roughly, I would say that 1.5x the shutter speed would cover it. If you were previously using 1/200th with a 100mm lens with a 16mp camera, then a 320th should be sufficient with a 22mp camera. Factor in the 2 stops of lattitude that IS provides and you're back to 1/80th. Considering the excellent performance of Canon DSLRs at high ISO (and I think there's no chance of them going backwards on this issue), an 80th is a very usable shutter speed for most lighting conditions, and if you think the performance of IS might be compromised with a higher pixel density camera and feel safer with a 1 stop advantage, then 160th would still be very usable.

Lastly, with many high tech systems which consist of separate components that develop at different rates (and Hi Fi systems are a typical example) it is not unusual for one part of the system to get ahead of another part such that the other part becomes the weakest link. When this happens, it is unreasonable to expect manufacturers of the stronger link to cease further development whilst waiting for the weakest link to catch up. We do not expect manufacturers of hi fi amplifiers to stop improving harmonic distortion figures because such improvements can not be fully reflected in even the best of loudspeakers.

Funnily enough, I was considering the same point after observing problems with hand-held shots, and I made a different suggestion.

I suggest that in addition to 1/(FL*CF) -- FL=focal length, CF=crop factor -- you need to take into account the pixel density. As a baseline, I propose using the pixel density of the 10D (ca. 8 microns per photosite). Let's call this the density factor (DF).

So, here's the New Truth (which seems to yield better results for me, anyway):

CF: crop factor (AKA FoV factor)
FL: focal length
DF: (sensor) density factor
PS: photosite size in microns
SS: shutter speed in seconds

DF = 8.0 / PS
SS = 1 / (FL * CF * DF)

For the EOS 20D/30D, we get:

SS = 1 / (FL * 1.6 * (8.0 / 6.4))
SS = 1 / (FL * 2)

For the D2x, we get:

SS = 1 / (FL * 1.52 * (8.0 / 5.5)
SS ~= 1 / (FL * 2.2)

Terminology experts: feel free to have a ball with those abbreviations.

I was tempted to include my original suggestion's clumsiness factor, but I guess everybody can add that on their own.

The problem with this approach, Jani, is that it's too technical and unecessarily precise for what is, after all, a rule-of thumb guideline. Even though I have some interest in the pixel pitch of my camera's sensor, I don't carry the precise micron measurement in my head. I'm aware that my 5D has a pixel pitch of around 8-9 microns and that the D60 (or 10D) is around 7 something microns etc, but to refresh my memory on this issue I'd have to do a Google search which, having just done it, would suggest that even you have got some of the figures slightly wrong in demonstrating your formula.

The pixel pitch of the 5D is 8.2 microns, that of the 10D 7.4 microns and the 30D 6.4 microns.

The 1/FL rule is based upon observations of 8x10" prints using 35mm film based cameras. For those not restricted to use of old-fashioned paper sizes, that's 8x12", which is as close as matters to the print size of a 5D at 240 ppi.

Now without getting into endless arguments about whether or not the 5D can resolve more than 35mm film, I'd suggest that the 5D's pixel density should be used as a baseline, bearing in mind that the resolving power of 35mm film varies considerably with the type of film used, Tech Pan and T-Max 100 being particularly good for resolution tests.

Using your formula from this baseline we get 1/(FL*8.2/6.4) for a full frame 22mp camera (or is that closer to 20mp? - I haven't got my calculator out). Whatever, it's a fairly marginal increase in shutter speed.

The real issue might be, is the 1/FL rule really adequate in the digital age? I suspect not. I'd prefer to amend it to a 1/2FL rule, for a hand-held shot, assuming the photographer is not suffering from Parkinson's, in which case it would be 1/4FL. If the photographer is into deep Yoga meditation and breathing exercises, then maybe 1/FL is still good enough .

Well, if it's too technical, then you could look at the introductory part and skip right to the numbers.

Seriously, though, using this formula helped me understand why doing what I'd done earlier with my 20D -- going for 1/FL or 1/(FL * 1.5) as the shutter speed for hand-held shots -- didn't work for me.

Going up to 1/(FL * 2) or quicker has helped with several shots.

That being said, there are circumstances where I've had successful hand-held shots at longer shutter speeds.

Your suggestion about going to 1/(FL * 2) as an approximation will work nicely with today's 8-12 Mpx cameras, but it's not future proof, and it doesn't work well with e.g. the Olympus E-x00 series.

Making a good rule of thumb that works for everybody isn't easy.

It's probably impossible. IS is supposed to give one a 2 stop advantage. The hype on the 24-105 zoom suggests 3 stops. I always discount that to one stop, especially with both long and short focal lengths. If I'm using my 100-400 IS zoom at 400mm, I want at least a 200th sec shutter speed. If I'm using my 24-105 IS zoom at 24mm, I won't even rely upon 1/13th sec for a sharp hand-held shot. I'd consider a 25th a minimum, although one can get lucky at 1/13th.

Perhaps I should say, I always try to discount that to 1 stop, but when using aperture priority it doesn't always work out that way .

And the hype on recent Nikkor lenses suggests 4 stops, with testing reports actually showing that it comes fairly close in at least some situations. Ditto for the EF 70-200, BTW, and 3 stops; it often matches and sometimes even exceeds a 3 stop advantage.


Yep, and I find myself using aperture priority most of the time. But it's nice to see IS working beyond expectations, too.

These formulae are good rules of thumb for practical action shooting, but if you are to take full advantage of the very high resulution that a 22 MP camera offers, I think that you really need mirror lockup and a sturdy tripod. We are talking about a resoltion that outresolves all but the best lenses and every article on lens testing I've ever read recommends at least a tripod.

A successful shot does not always require the full resolution of which the camera is capable. Some testing with Imitest or a resolution chart would give some idea of the resolution one can achieve with hand holding and vibration reduction technology.

Bjanes,
You may not be aware, but there have been many tests that have demonstrated that MLU is only of benefit for a fairly narrow range of slow shutter speeds, typically between about 2 seconds and 1/30th. It may extend slightly from 4 seconds to 1/60th, but that's about it.

I have no information on tripod shots at slow shutter speeds versus hand-held shots at high shutter speeds. Understandably, that would be a highly variable situation, depending upon tripod stability and photographer stability.

At what point does all this become more hurtful than good? Much of this talk about 22 MP sensors and 2 and 3 times the shutter speed is all good and fine, but is quite worthless to many photographers. Aside from tripod and cable release toting photographers (of which I am one, from time to time) who is going to want this course of action?

I think you've misunderstood the point. 2 & 3 times the shutter speed is not a recommendation as a result of increased pixel count. Anyone who is concerned about getting the maximum resolution from his/her camera should be aware that adequate shutter speed (or use of a tripod) is critical to achieving this.

Any sensor with increased pixel count will simply 'up the ante' with regard to optimal shutter speed for maximum resolution, but not by much.

Ray, because digital allows cost free shooting I spent a rainy day firing off hundreds of shots to understand the relationship between shutter speed, mirror lock up, hand-holding, and tripod weight. I used a range of Hasselblad cameras (503CW, 555ELD, 903 CW/F) and a Canon 1Ds MkII, the target was a light bulb within a box, covered by tin foil with tiny pin pricks to record just points of light.

I'm afraid that for me and my equipment the benefit of mirror lock up extends to much faster speeds than 1/60s.

Using a Hasselblad with a P25 back MLU was giving a noticeable sharpness benefit all the way to the top speed of 1/500s. This applied to both small and large Gitzo carbon fibre tripods, however I also used a massive studio tripod (far too massive to use outside of a studio) and okay there the MLU benefit was smaller and seemed to top out at about 1/60s or 1/125s.

What was scary was just how sensitive the whole process was. Handholding the Canon, even with fairly high shutter speeds and Image Stabilising lenses still showed degradation of the image. The results from hand-holding the Hasselblad 503 CW would have been comical but for the fact that I've spent thirty years hand-holding Hasselblads!

Furthermore, just using a tripod doesn't guarantee success. The light Gitzo carbon fibre tripod was better than hand-holding, but in turn it wasn't nearly as good as the heavy Gitzo carbon fibre tripod. And at speeds under about 1/30s there were further small benefits from the massive studio tripod.

The exercise was sobering in that it demonstrated to me that chasing the last 5 or 10% of photographic quality can be a fool's errand, unless you're prepared to go to really quite extraordinary and impractical lengths. It also explained why, looking back over a lifetime of photography, my sharpest shots have consistently been achieved in a studio environment with huge tripods and high speed strobes.

Gary,
That's interesting, and as you say, a bit scary. The 4 sec to 1/60th danger area that benefits from MLU probably refers to 35mm film tests. My own tests were done some time ago using a D60 on various tripods ranging from fairly sturdy to quite flimsy and seemed to confirm this range of shutter speeds, provided the tripod used was sturdy with no disturbing influences such as even a slight breeze. With a tripod that was really too light for the load, or a tripod outside in slightly breezy conditions, I recall there was no point at all in MLU whatever the shutter speed.

Another factor which is going to influence the result is the dampening of the mirror. It would be unreasonable to assume that all cameras have equally dampened mirror slap, so perhaps here we are into a situation where there's no general rule that will apply across all formats and camera models.

Perhaps I should repeat such tests with my 5D. I'm a firm believer in knowing the qualities and limitations of one's own equipment.